US9721661B1 - Content addressable memories - Google Patents

Content addressable memories Download PDF

Info

Publication number
US9721661B1
US9721661B1 US15/216,011 US201615216011A US9721661B1 US 9721661 B1 US9721661 B1 US 9721661B1 US 201615216011 A US201615216011 A US 201615216011A US 9721661 B1 US9721661 B1 US 9721661B1
Authority
US
United States
Prior art keywords
voltage
search
match
memristor
bit
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
US15/216,011
Inventor
Brent Buchanan
Le Zheng
John Paul Strachan
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hewlett Packard Enterprise Development LP
Original Assignee
Hewlett Packard Enterprise Development LP
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hewlett Packard Enterprise Development LP filed Critical Hewlett Packard Enterprise Development LP
Priority to US15/216,011 priority Critical patent/US9721661B1/en
Assigned to HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP reassignment HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BUCHANAN, BRENT, STRACHAN, JOHN PAUL, ZHENG, Le
Priority to KR1020170092234A priority patent/KR20180011016A/en
Application granted granted Critical
Publication of US9721661B1 publication Critical patent/US9721661B1/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C15/00Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C15/00Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores
    • G11C15/04Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores using semiconductor elements
    • G11C15/046Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores using semiconductor elements using non-volatile storage elements
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/903Querying
    • G06F16/90335Query processing
    • G06F16/90339Query processing by using parallel associative memories or content-addressable memories
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C11/00Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
    • G11C11/02Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using magnetic elements
    • G11C11/16Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using magnetic elements using elements in which the storage effect is based on magnetic spin effect
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C13/00Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00
    • G11C13/0002Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C13/00Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00
    • G11C13/0002Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements
    • G11C13/0007Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements comprising metal oxide memory material, e.g. perovskites
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C15/00Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores
    • G11C15/04Digital stores in which information comprising one or more characteristic parts is written into the store and in which information is read-out by searching for one or more of these characteristic parts, i.e. associative or content-addressed stores using semiconductor elements

Definitions

  • CAM Content addressable memory
  • search operation in which a data string may be input as search content and the resulting output is an address of a location in the memory that stores matching data (if there is any). This is in contrast to a read operation in which an address is input and the resulting output is the data stored in the memory location corresponding to the searched address.
  • Certain CAMs may be able to perform both the aforementioned search operation and the aforementioned read operation, while non-CAM memories may be able to perform the read operation but not the search operation.
  • FIG. 1A is circuit diagram illustrating an example bit cell of an example CAM.
  • FIG. 1B is a circuit diagram illustrating an example resistive divider.
  • FIG. 2 is a table illustrating an association between resistance states and stored values.
  • FIG. 3 is a table illustrating search results for combinations of search criteria and stored value.
  • FIG. 4 is a table illustrating voltages that are applied to wiring lines of an example CAM to perform writing and reading operations.
  • FIG. 5A is a table illustrating voltages that are applied to wiring lines of an example CAM to perform search operations.
  • FIG. 5B is a signal diagram illustrating voltages that are applied to wiring lines of an example CAM to perform search operations, as well as the states of an example bit cell resulting from such search operations for each combination of search criterion and stored value.
  • FIG. 6 is a diagram illustrating an example CAM with an array of multiple bit cells.
  • FIG. 7 is a diagram illustrating an example CAM with an array of bit cells storing particular values.
  • FIG. 8 is a diagram illustrating an example device that includes an example CAM as a component thereof.
  • CAMs that include memristors as storage elements. Also described herein are example devices that may have a CAM as a component thereof, as well as example methods of operating CAMs.
  • the example CAMs described herein may include memristors as storage elements.
  • the example CAMs disclosed herein may have higher storage densities and lower power consumption than comparable CAMs that use SRAMs or DRAMs as storage elements.
  • certain of the example CAMs disclosed herein may be capable of being searched on the basis of three search criteria (e.g., 0, 1, and a wildcard), which may provide increased flexibility and speed in search operations as compared to a CAM that is searchable on the basis of two search criteria (e.g., 0 and 1).
  • three search criteria e.g., 0, 1, and a wildcard
  • the example CAMs disclosed herein may have bit cell circuit layouts that can fit in a smaller layout space than other CAM bit cell layouts, which may allow the example CAMs to have a more densely packed memory array (i.e., more bit cells per unit area) than other comparable CAMs.
  • bit cell layout space might be achieved, for example, as a result of the bit cell circuit layout allowing for the use of just two FETs per bit cell instead of the three or more FETs per bit cell that may be required in other circuit layouts in order to provide CAM functionality.
  • all of the transistors in the bit cells of the CAM may be a same channel type as one another (for example, all NFETs); this may allow for a well to be omitted from the bit cell array that might otherwise be needed, thus freeing up layout space that would have been used for the well.
  • the use of transistors having all the same channel type may also decrease the complexity and/or cost of manufacturing the CAM.
  • a bit cell may include a memristor, a switching transistor, and a match-line transistor.
  • the memristor and the switching transistor may be connected in series between a first data line and a second data line.
  • the match-line transistor may be connected between a match-line and a rail.
  • a gate of the match-line transistor may be connected to a common node of the memristor and the switching transistor.
  • the switching transistor may be sized such that its channel resistance when on (i.e., when ohmic) is between a resistance associated with a low-resistance state of the memristor and a resistance associated with a high-resistance state of the memristor.
  • a gate of the switching transistor may be connected to a word line.
  • certain of the example CAMs may also include control circuitry that searches the bit cell based on a first search criterion (e.g., 0 ), based on a second search criterion (e.g., 1 ), and based on a wildcard search criterion. If the bit cell stores a value that does not match the search criterion, the voltage of the common node during the search may become a voltage that turns on the match-line transistor, thereby connecting the match line to the rail and resulting in a voltage of the match line (which has been pre-charged) being pulled down below its pre-charged voltage.
  • a first search criterion e.g., 0
  • a second search criterion e.g., 1
  • a wildcard search criterion e.g. 1
  • the voltage of the common node may become a voltage that keeps the match-line transistor off so that the voltage of the match-line is not pulled down (at least not by that bit cell—it might be pulled down by some other bit cell).
  • matches or misses of the search criterion may be determined by monitoring the match line and determining whether or not its voltage is pulled low.
  • the control circuit may search the bit cell based on the first search criterion by pre-charging the match line, applying a first voltage to the first data line, applying a ground voltage to the second data line, and applying a third voltage to the rail.
  • the control circuit may search the bit cell based on a second search criterion by pre-charging the match line, applying the ground voltage to the first data line, applying the first voltage to the second data line, and applying a fourth voltage to the rail.
  • the control circuit may search the bit cell based on the wildcard search criterion by applying a ground voltage to both the first and second data lines.
  • the memristor and the switching transistor of the bit cell may form a resistive divider.
  • An output voltage of the resistive divider i.e., the voltage of the common node between the memristor and the switching transistor
  • the output voltage of the resistive divider may control whether the match-line transistor is on or off.
  • the relative resistances of the memristor and the switching transistor and the voltages that are applied to the bit cell during a search operation may be set such that the output voltage of the resistive divider during the search operation is an off-voltage (low voltage) when the memristor stores a value that matches the search criterion and an on-voltage (high voltage) when the memristor stores a value that matches the search criterion.
  • the search criterion is the wildcard search criterion
  • the output voltage of the resistive divider is low regardless of what value is stored in the resistive divider.
  • Certain example CAMs may include multiple instances of the example bit cells described above, which may be arrayed in rows and columns.
  • each first and second data line may be connected to each bit cell in a same column, and each word line may be connected to each bit cell in a same row.
  • Each match line may be connected to multiple bit cells that are in a same row as one another, with the number of bit cells per match line being equal to a search unit size for the CAM (i.e., a word size) and the bit cells that are connected to the same match line as one another forming a word storage block.
  • the bit cells of the word storage block may be searched to see if the data stored in the word bit cells of the word storage block matches an input search word.
  • the match line for a given word storage block may be pulled low if any one of the bit cells thereof misses its respective search criterion, and therefore it can be determined whether the given word storage block stores data that matches the search word by monitoring whether the corresponding match line is pulled low.
  • Each row may include a single word storage block (in which case there may be one word line per row), or each row may include P>1 word storage blocks (in which case there may be P match lines per row).
  • FIG. 1A illustrates an example bit cell 110 of an example CAM.
  • the example bit cell 110 may include a memristor M, a transistor T 1 , and a match-line transistor T 2 .
  • the memristor M may be connected between a first data line DL 1 and a node G.
  • the transistor T 1 may be connected between the node G and a second data line DL 2 .
  • the memristor M and the transistor T 1 may be connected via a common node, namely the node G.
  • the match-line transistor T 2 may be connected between a rail RL and a match line ML.
  • a gate of the match-line transistor T 2 may be connected to the node G, and therefore a voltage of the node G may control a conductive state of the match-line transistor T 2 .
  • a gate of the transistor T 1 may be connected to a word line WL, and therefore a voltage of the word line WL may control a conductive state of the transistor T 1 .
  • the memristor M may serve as the storage element of the bit cell 110 .
  • the memristor M may be a device whose resistance can be changed between multiple resistance states by applying certain voltage differences across the memristor M or flowing certain currents through the memristor M, with the memristor M “remembering” or maintaining its most recent resistance state even when the voltage/current that caused it to enter that state has been removed.
  • the current resistance state of the memristor M may depend on the magnitude and polarity of voltage differences that have been applied across the memristor in the past or on how much current has flowed in what direction through the memristor M in the past.
  • the memristor M may be capable of changing between a low resistance state in which the memristor M exhibits relatively low resistance and a high resistance state in which the memristor M exhibits relatively high resistance.
  • the memristor M may be caused to change to the low resistance state by applying a setting voltage difference of sufficient magnitude with a first polarity across the memristor, and the memristor M may be caused to change to the high resistance state by applying a resetting voltage difference of sufficient magnitude across the memristor M with a second polarity.
  • the resistance state of the memristor M may be used to store data values.
  • a high resistance state of the memristor M may represent a first data value and a low resistance state of the memristor M may represent a second data value.
  • a high resistance state represents a digital 0 and a low resistance state represents a digital 1 , as illustrated in the table of FIG. 2 ; however, it should be understood that the logical association between resistance state and stored value may be arbitrarily selected.
  • the memristor M may be set to a low resistance state by turning on the switching transistor T 1 (e.g., apply V DD to the word line WL), applying the setting voltage V set to the first data line DL 1 , and applying a ground voltage GND to the second data line DL 2 , as illustrated in FIG. 4 .
  • the memristor M may be reset to a high resistance state by turning on the switching transistor T 1 (e.g., apply V DD to the word line WL), applying the resetting voltage V reset to the second data line DL 2 , and applying the ground voltage GND to the first data line DL 1 , as illustrated in FIG. 4 .
  • V set and V reset may be set to any values that are larger than minimum values required to be able to set and reset the memristor M, respectively.
  • V set _ min is the minimum voltage that will result in setting the memristor M
  • V reset _ min is the minimum voltage that will result in resetting the memristor M
  • V set may be equal to or greater than V set _ min
  • V reset may be equal to or greater than V reset _ min .
  • V set and V reset do not necessarily need to have the same magnitude, since V set _ min is not necessarily the same as V reset _ min .
  • V set and V reset could be set to the same value as long as that value is greater than or equal to both V set _ min and V reset _ min .
  • V DD is a highest voltage level of the TCAM.
  • the values of V set _ min and V reset _ min may be parameters of the memristor M that are known in advance, that are determined theoretically based on a model of the memristor M, or that are determined experimentally.
  • the value that is stored in the bit cell 110 may be read by applying a read voltage difference across the memristor M and measuring a current that flows through the memristor M.
  • the magnitude of the current that flows through the memristor M may be used to determine the resistance state of the memristor M because a high resistance state will result in low current, while a low resistance state will result in high current.
  • the value stored in the bit cell may be read by turning on the switching transistor T 1 (e.g., apply V DD to the word line WL), applying the read voltage V read to either one of the first and second data lines DL 1 /DL 2 , and applying the ground voltage GND to the other one of the first and second data lines DL 1 /DL 2 , as illustrated in FIG.
  • the current that flows through the memristor M may be sensed on either one of the data lines DL 1 /DL 2 , or the current flowing on the match line ML may be sensed.
  • applying voltage differences across the memristor M whose magnitudes are relatively small in comparison to the setting/resetting voltage differences will not significantly affect the resistance state of the memristor M.
  • the read voltage difference may be set to a value that is below a minimum voltage that will set or reset the memristor M.
  • the resistance state of the memristor M may be determined by measuring a voltage of the match line ML during a reading operation.
  • the match-line transistor T 2 may be set to act as an amplifier (e.g., common source amplifier or source-follower amplifier), and the voltage of the match line ML may vary depending on the voltage of the node G. Since the voltage of the node G during a reading operation depends on the resistance state of the memristor, the voltage of the match line ML will also depend on the resistance state of the memristor.
  • the bit cell 110 may be thought of as including a resistive divider 111 , with the memristor M and the switching transistor T 1 serving as the resistors of the resistive divider 111 .
  • a resistive divider is formed when two resistors are connected in series, input voltages are applied to opposite input terminals of the resistors, and an output voltage is taken from a common node between the resistors.
  • FIG. 1 FIG. 1
  • FIG. 1B illustrates a generalized form of a resistive divider, with resistors R 1 and R 2 being connected in series, input voltages V IN _ 1 and V IN _ 2 being applied to the input terminals, and the output voltage V OUT being taken from the common node between the resistors R 1 and R 2 (which may be referred to as the output terminal).
  • the output voltage of a generalized resistive divider may be given by the following:
  • V OUT R 2 ⁇ V IN ⁇ ⁇ _ ⁇ ⁇ 1 + R 1 ⁇ R IN ⁇ ⁇ _ ⁇ ⁇ 2 R 1 + R 2 ( eq . ⁇ 1 )
  • the memristor M and the switching transistor T 1 may be considered as being the resistors of the resistor divider 111 , and correspond, respectively, to the resistors R 1 and R 2 of the generalized resistive divider.
  • a memristor is not the same type of circuit component as a resistor, as the memristor can have its resistance changed over time while a resistor generally has a constant resistance.
  • the memristor has an approximately constant resistance, and thus the memristor M may be modeled approximately during these periods as a resistor.
  • the switching transistor T 1 may be sized such that it has a non-negligible channel resistance when the transistor T 1 is on (i.e., when it is ohmic), and this channel resistance may be modeled as a resistor.
  • the switching transistor T 1 may be sized such that its channel resistance when on (R T1 ) is greater than a resistance associated with a low resistance state of the memristor M (R LRS ), but less than a resistance associated with a high resistance state of the memristor M (R HRS )—that is, R LRS ⁇ R T1 ⁇ R HRS .
  • R LRS low resistance state of the memristor M
  • R HRS a resistance associated with a high resistance state of the memristor M
  • the two input terminals thereof are formed from the terminal of the memristor M that is connected to the first data line DL 1 and the terminal of the transistor T 1 that is connected to the second data line DL 2 .
  • the input voltages of the resistive divider 111 are the voltage carried by the first data line DL 1 (V DL1 ) and the voltage carried by the second data line DL 2 (V DL2 ).
  • the input voltages V DL1 and V DL2 of the resistive divider 111 correspond, respectively, to the input voltage V IN _ 1 and V IN _ 2 of the generalized resistive divider.
  • the output terminal of the resistive divider 111 is the node G, and the output voltage of the resistive divider 111 is the voltage of the node G (V G ), which corresponds to the output voltage V OUT of the generalized resistive divider.
  • V G the voltage of the node G when the transistor T 1 has been turned on
  • V G R T ⁇ ⁇ 1 ⁇ V D ⁇ ⁇ L ⁇ ⁇ 1 + R M ⁇ R D ⁇ ⁇ L ⁇ ⁇ 2 R M + R T ⁇ ⁇ 1 ( eq . ⁇ 2 ) where R M is the present resistance of the memristor M.
  • the example bit cell 110 may be searched based on three search criteria—namely 0, 1, and a wildcard (denoted Y herein for convenience).
  • the configuration of the bit cell 110 and the voltages that are applied to the bit cell during a search operation are set such that search operations produce the search results illustrated by the table in FIG. 3 .
  • the result is a match if the bit cell 110 stores a 0 (i.e., if R M ⁇ R HRS ) and a miss if the bit cell 110 stores a 1 (i.e., if R M ⁇ R LRS ).
  • the result is a miss if the bit cell 110 stores a 0 (i.e., if R M ⁇ R HRS ) and a match if the bit cell 110 stores a 1 (i.e., if R M ⁇ R LRS ).
  • the bit cell 110 is searched based on the wildcard search criterion (Y)
  • the result is a match regardless of what value is stored by the bit cell 110 .
  • FIGS. 5A and 5B illustrate how search operations based on the above-noted search criteria may be performed, as well as states of the bit cell in response to such search operations for each combination of search criteria and stored value.
  • the match line ML may be pre-charged to a pre-charge voltage V pre
  • the first transistor T 1 may be turned on by applying V DD to the word line WL
  • a search voltage V search may be applied to the first data line DL 1 while the ground voltage GND is applied to the second data line DL 2
  • a voltage V 0 may be applied to the rail RL, as illustrated in FIGS. 5A and 5B .
  • the match line ML may be pre-charged to a pre-charge voltage V pre
  • the first transistor T 1 may be turned on by applying V DD to the word line WL
  • the ground voltage GND may be applied to the first data line DL 1 while the search voltage V search is applied to the second data line DL 2
  • a voltage V 1 may be applied to the rail RL.
  • the voltage of the node G may control a conductive state of the match-line transistor T 2 —in particular, the match-line transistor T 2 may be turned on when a gate-source voltage thereof is greater than a threshold voltage thereof (V th ), and may be kept off when a gate-source voltage thereof is less than V th .
  • V th a threshold voltage thereof
  • the voltage of the node G during a search operation will depend on the resistance state of the memristor M.
  • the conductive state of the match-line transistor T 2 during a search operation may be controlled based on the resistance state of the memristor M. In particular, by appropriately setting the various resistances of the bit cell 110 and the voltages applied thereto during the search operation, it can be assured that the search results illustrated in FIG. 3 are obtained.
  • V 0 the gate-source voltage of the match-line transistor T 2 may be kept below the threshold voltage V th , and therefore the match-line transistor T 2 may be kept off.
  • the bit cell 110 does not pull the voltage of the match line ML down, thereby indicating a match.
  • V 1 the gate-source voltage of the match-line transistor T 2 in this state may be raised above the threshold voltage V th , and the match-line transistor T 2 may be turned on.
  • the match line ML is connected to the rail RL and begins to discharge, thus pulling the voltage of the match line ML low (thereby indicating a miss).
  • the voltages V 0 , V 1 , and V search , the resistances R T1 , R HRS , R LRS , and the threshold voltage V th may be set such that:
  • the resistances R T1 , R HRS , R LRS and/or the threshold voltage V th may be set based on other considerations, in which case the conditions of equation 9 may be satisfied by setting V 0 , V 1 , and V search .
  • V search may be less than V set _ min and less than V reset _ min , so that the search operation does not inadvertently set or reset the memristor M.
  • V search may be the same value as V read , since V read is less than V set _ min and V reset _ min .
  • Using the same voltage for V read and V search may reduce the complexity of drivers that are used to supply the voltages to the bit cell 110 , as fewer distinct voltages may need to be generated.
  • the match line ML may be pre-charged to a pre-charge voltage V pre
  • the first transistor T 1 may be turned on by applying V DD to the word line WL
  • the ground voltage GND may be applied to the first data line DL 1 , the second data line DL 2 , and the rail RL.
  • the voltage of the node G during a search based on the Y search criterion becomes GND, regardless of what value is stored in the bit cell 110 . Because the voltage of the node G during a Y search is always GND and because GND is applied to the rail RL. the match-line transistor T 2 is kept off (indicating a match) regardless of the value stored in the bit cell 110 .
  • FIG. 6 illustrates an example CAM 100 that includes an array 140 of multiple bit cells 110 .
  • the bit cells 110 may be arrayed in W rows and N columns.
  • each row corresponds to a word storage block (e.g., word 0 through word W ⁇ 1), with each bit cell 110 in a given row corresponding to a bit position (e.g., bit 0 through bit N ⁇ 1).
  • the CAM 100 may include multiple word lines WL (e.g., WL 0 though WL W-1 ), multiple match lines ML (e.g., ML 0 though ML W-1 ), multiple first data lines DL 1 (e.g., DL 1 0 though DL 1 N-1 ), multiple second data lines DL 2 (e.g., DL 2 0 though DL 2 N-1 ), and multiple rails RL (e.g., RL 0 through RL N-1 ).
  • Each bit cell 110 in a given row may be connected to the same word line WL and to the same match line ML as one another.
  • all of the bit cells 110 in a given word may be selected at the same time for writing/reading by turning on the word line WL that corresponds to the given word.
  • the match line ML of a given word may be pulled low if any one of the bit cells 110 that correspond to the word is a miss.
  • Each bit cell 110 in a given column may be connected to the same first data line DL 1 , the same second data line DL 2 , and the same rail RL as one another. Thus, each bit cell 110 in a same column may be searched simultaneously based on the same search criterion.
  • the CAM 100 may include control circuitry 150 that applies signals to the various lines mentioned above so as to control operations of the bit cells, such as read, write, and search operations.
  • the control circuitry 150 may comprise drivers, shift registers, decoders, and/or the like that are connected to the various wiring lines of the array 140 .
  • the control circuitry 150 may receive a control signal Ctrl that instructs the control circuitry 150 which operation it should perform.
  • the control circuitry 150 may also receive a read/write address that indicates a target word storage block for a read operation or a write operation.
  • the control circuitry 150 may output an output word as a result of performing a read operation, with the output word corresponding to the word that is stored in the word storage block whose address corresponds to the read/write address.
  • the control circuitry 150 may also receive an input word that indicates either a word that the CAM 100 is to be searched for during a search operation or a word that is to be written to a particular word storage block during a write operation.
  • the control circuitry 150 may also output an output address as a result of performing a search operation, with the output address corresponding to the address (or addresses) of any word storage block(s) whose stored word matches the input search word.
  • the control circuitry 150 is illustrated as including a row driver 151 and a column driver 152 .
  • the row driver 151 is connected to the word lines WL 0 though WL W-1 and the match lines ML 0 though ML W-1 , and may supply electrical signals to and/or sense electrical signals carried on these wiring lines.
  • the column driver 152 is connected to the first data lines DL 1 0 though DL 1 N-1 , the second data lines DL 2 0 though DL 2 N-1 , and the rails RL 0 through RL N-1-1 , and may supply electrical signals to and/or sense electrical signals carried on these wiring lines.
  • control circuitry 150 is merely for ease of description, and that it does not imply that the control circuitry 150 is necessarily composed of exactly two drivers in the configuration illustrated.
  • the control circuitry 150 could include a separate driver for each type of wiring line, a single driver for all the wiring lines, or any combination of drivers.
  • the illustrated locations of the drivers relative to the array 140 are not intended to be meaningful, and it should be understood that drivers could be located on any side or combination of sides of the array 140 .
  • the control circuitry 150 may control the CAM 100 to perform a read operation and a write operation in units of words.
  • the row driver 151 may select the target word storage block based on the input read/write address by applying V DD to its corresponding word line WL, and the column driver 152 may then proceed to apply voltages to the column lines to perform the read/write operations for each bit cell 110 in the selected word storage block.
  • the column driver 152 may apply voltages to each of the first and second data lines DL 1 /DL 2 and rails RL according to the table illustrated in FIG. 4 .
  • the current flowing through each of the N bit cells 110 of the target word storage block may be sensed by the column driver 152 (for example, via either the first or second data lines DL 1 /DL 2 ).
  • the column driver 152 may determine a bit value stored in each bit cell 110 based on the current that flows therethrough, and these N bit values may be combined into the output word that is output from the column driver 152 .
  • the column driver 152 may determine, based on the input word, whether each of the N bit cells 110 in the target word storage block is to be set or reset and, in accordance with this determination, set or reset each of the N bit cells 110 in the target word storage block by applying the appropriate voltages to the corresponding first and second data lines DL 1 /DL 2 and rail RL.
  • setting and resetting operations may be performed simultaneously across the target word storage block.
  • setting and resetting operations may be performed in distinct phases. For example, all of the bit cells 110 in the target word storage block may first be reset, and then those bit cells 110 that are to be set may subsequently be set.
  • bit cells 110 could be written to or read from during a read or write operation without necessarily writing to or reading the other bit cells 110 in the same word storage block.
  • the word line WL of the word storage block that contains the target bit cell(s) 110 could be selected, the ground voltage GND could be applied to all of the first and second data lines DL 1 /DL 2 that correspond to bit cells 110 that are not the target bit cell(s), and the voltages for the writing/reading operation could be applied to the first and second data lines DL 1 /DL 2 of the target bit cell(s) 110 .
  • any bit cells 110 that are not desired to be writing to or read from can be omitted from the writing/reading operation simply by applying GND to their corresponding data lines DL 1 /DL 2 .
  • the control circuitry 150 may control the CAM 100 to perform a search operation across the entire array 140 simultaneously.
  • the row driver 151 may select all of the target word storage blocks simultaneously by applying V DD to all of the word lines WL, and may pre-charge all of the match lines ML.
  • the column driver 152 may then proceed to apply voltages to the column lines to perform the search operations for each bit position based on the input search word.
  • the column driver 152 may determine, based on the input search word, which search criterion is to be used to search each bit position, and in accordance with this determination, apply the appropriate voltages to the corresponding first and second data lines DL 1 /DL 2 and rail RL.
  • the row driver 151 may then monitor the match lines to determine which are pulled low (indicating the corresponding word storage block stores a word that is a miss) and which remain charged (indicating the corresponding word storage block stores a word that is a match).
  • the row driver 151 may decode the results of misses/matches from the match lines ML and output an address (or addresses) of a matching word storage block (or matching words storage blocks).
  • FIG. 7 which includes an array 140 having three rows and three columns of bit cells 110 , with the bit cells 110 each storing values as indicated in the figure.
  • this example CAM 100 is searched with the input search word being “Y 10 ”.
  • the column driver 152 will receive the input search word and the control signal Ctrl indicating a search operation, and based thereon would determine that the bit- 0 column should be searched based on the wildcard search criterion (since the 0th bit of the search word is Y), the bit- 1 column should be searched based on the 1 search criterion (since the 1st bit of the search word is 1), and the bit- 2 column should be searched based on the 0 search criterion (since the 2nd bit of the search word is 0). Accordingly, the column driver 152 may perform the search operation by applying voltages to the various wiring lines as follows:
  • the bit- 0 column Since the bit- 0 column is searched on the basis of the wildcard criterion, none of the bit cells 110 in this column will be a miss, and hence none of these bit cells 110 will pull down their respective match lines ML.
  • the voltage of the node G in each bit cell 110 in the bit- 0 column will be GND because GND is applied to both DL 1 0 and DL 2 0 . Because GND is also applied to RL 0 , the gate-source voltage of the match-line transistor T 2 of each bit cell 110 in the bit- 0 column will be zero, and hence the match-line transistor T 2 of each bit cell 110 in the bit- 0 column will remain off.
  • the [bit- 1 , word- 0 ] bit cell 110 stores a 1 and is searched on the basis of the 1 search criterion, and hence from equation 8 the voltage of node G becomes
  • the [bit- 2 , word- 0 ] bit cell 110 stores a 0, and is searched on the basis of the 0 search criterion, and hence from equation 5 the voltage of node G becomes
  • the row driver 151 determines that the word 0 is a match for the search word. Thus, an address of the word 0 is output from the row driver 151 as an output address.
  • the [bit- 1 , word- 1 ] bit cell 110 stores a 0 and is searched on the basis of the 1 search criterion, and hence from equation 6 the voltage of node G becomes
  • the [bit- 2 , word- 1 ] bit cell 110 stores a 1 and is searched on the basis of the 0 search criterion, and hence from equation 7 the voltage of the node G becomes
  • the row driver 151 determines that the word 1 is a miss for the search word. Thus, an address of the word 1 is not output from the row driver 151 as an output address.
  • the states of the [bit- 1 , word- 2 ] and [bit- 2 , word- 2 ] are similar to the states of the [bit- 1 , word- 0 ] and [bit- 2 , word- 0 ] bit cells 110 that were described above, and hence they will not be described in detail.
  • the [bit- 1 , word- 2 ] and [bit- 2 , word- 2 ] bit cells 110 are matches, neither one pulls down the match line ML 2 , and therefore the row driver 151 determines that the word 2 is a match for the search word.
  • an address of the word 2 is output from the row driver 151 as an output address.
  • the example CAMs described herein may be used in any electronic device in which a memory might be useful, such as, for example, in a personal computer, server, smartphone, tablet, network device, etc.
  • the example CAMs can be used to store data for the device by writing the data into the CAM, and the device can access the stored data by reading from the CAM.
  • the example CAMs may also allow the device to search the CAM based on an input search word, which can be particularly useful in a variety of applications.
  • a database application executed by an example device may benefit greatly by the device using the example CAMs.
  • a network device (such as a router, switch, gateway, etc.) may greatly benefit from using the example CAMs, as the high search speed provided by the CAM may allow the network device to operate at higher speeds than may otherwise be possible.
  • FIG. 8 illustrates an example device 500 that includes an example CAM 100 .
  • the example device 500 could be any type of electronic device.
  • some of the features described below could be omitted from the example device 500 and others not described below could be added.
  • the device 500 may include a CAM 100 , a processing circuitry 510 , communications circuitry 520 , and machine readable media 530 .
  • the CAM 100 may be an instance of the example CAM 100 described above, and may include an array 140 of bit cells 110 and control circuitry 150 .
  • the processing circuitry 510 may supply the control signal Ctrl to the control circuitry 150 .
  • the processing circuitry 510 may also supply the input word and the read/write address to the control circuitry 150 .
  • the processing circuitry 510 may be any circuitry capable of executing machine-readable instructions, such as a central processing unit (CPU), a microprocessor, a microcontroller device, a digital signal processor (DSP), etc.
  • the processing circuitry 510 may also be an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), an application-specific instruction set processor (ASIP), or the like, that is configured to perform certain operations described herein, such as the read, write, and search operations.
  • ASIC application-specific integrated circuit
  • FPGA field-programmable gate array
  • ASIP application-specific instruction set processor
  • the machine readable media 530 may be any non-transitory machine readable medium, which may include volatile storage media (e.g., DRAM, SRAM, etc.) and/or non-volatile storage media (e.g., PROM, EPROM, EEPROM, NVRAM, hard drives, optical disks, etc.).
  • volatile storage media e.g., DRAM, SRAM, etc.
  • non-volatile storage media e.g., PROM, EPROM, EEPROM, NVRAM, hard drives, optical disks, etc.
  • the machine readable media 530 may store machine-readable instructions that, when executed by the processing circuitry 510 , cause the device 500 to perform some or all of the operations described herein, such as the read, write, and search operations.
  • the communications circuitry 520 may be circuitry for receiving input data communications and sending output data communications.
  • the communications circuitry 520 may include a network interface card.
  • the communications circuitry 520 may include multiple communications ports 521 , and may serve to connect multiple other electronic devices to one another via the device 500 .
  • the device 500 may be a network router, network switch, network gateway, or the like, and may perform various traffic control tasks such as routing, switching, etc.
  • the communications circuitry 520 may receive communications that include a destination address (such as an IP address), and the communications circuitry 520 may determine which communications port 521 to forward the received communication to based on the destination address.
  • the communications circuitry 520 may determine which port 521 to forward a received communication to by searching the CAM 100 based on the destination address that was included in the communication.
  • each device that is connected to the device 500 may have a unique device address (such as an IP address), and the device 500 may store the device address of each device that is connected thereto in the CAM 100 such that the location of the stored device address within the CAM 100 corresponds to the communications port 521 to which the device is connected.
  • each word storage block of the CAM 100 may correspond to a different port 521 of the device 500 , and whenever a new device is connected to the device 500 the device address of the new device may be written into a word storage block that corresponds to the port 521 to which the new device is connected.
  • the destination address of the communication may be sent to the CAM 100 (either directly, or via the processing circuitry 510 ) as an input search word; the CAM 100 may then return a memory address of a matching word storage block to the communications circuitry 520 (either directly or via the processing circuitry 510 ).
  • the memory address of the word storage block may be understood by the communications circuitry 520 to identity a particular port 521 , and therefore the communications circuitry 520 can determine which port 521 to forward the communication to based on the output address of the CAM 100 .
  • the memory address of each word storage block may be the same as the address of one of the ports 521 , which may enable the communications circuitry 520 to easily determine from an output memory address which port 521 corresponds thereto.
  • the CAM 100 may operate as sort of look-up table that indicates which port corresponds to which destination address. However, because the entire array 140 of the CAM 100 may be searched simultaneously and may be searched at the hardware level, the amount of time it takes for the device 500 to determine which port to forward a communication to can be greatly reduced.
  • the example memristor M is a bipolar memristor—i.e., it is set by a voltage difference/current having one polarity and reset by a voltage difference/current having a different polarity.
  • the memristor M may be a unipolar memristor—i.e., a memristor that is set and reset by the same polarity.
  • unipolar memristors the same reading and searching operations that were described above with regard to a bipolar memristor could be used, and any known unipolar memristor writing method could be used in place of the writing method described above.
  • the memory element of the bit cell 110 is described as a memristor.
  • the term “memristor” may be used in certain contexts in a broad sense and may be used in certain contexts in a narrow sense.
  • “memristor” may refer specifically to circuit elements that exhibit a non-linear relationship between electric charge and magnetic flux (or exhibit a relationship between the time integral of current and the time integral of voltage).
  • “memristor” refers broadly to any non-volatile memory element that is based on changing resistance states and is read by its resistance.
  • memristor is always used in the broad sense, unless specifically indicated otherwise; specifically, as used herein, “memristor” refers broadly to any non-volatile memory element that is based on changing resistance states and is read by its resistance.
  • the memristor M may be, for example, a memristor in the narrow sense of the term, a memory element that changes the resistance across a dielectric solid-state material, a phase-change memory element (sometimes referred to as PCM or PCRAM), a conductive-bridging memory element (sometimes referred to as CBRAM or a programmable metallization cell), a magnetoresistive memory element (sometimes referred to as MRAM), or the like.
  • circuit components are conceptual representations of classes of physical structures or devices that perform certain functions and/or have certain properties. Examples of such circuit components include passive devices such as resistors, capacitors, memristors, etc.; active devices such as transistors, diodes, etc.; constituent elements of the active/passive devices such as terminals, electrodes, gates, sources, drains, etc.; elements that connect devices such as wiring lines, nodes, etc.; and so on.
  • a single physical structure (or set of physical structures) in an actual physical incarnation of an example CAM may serve multiple functions and/or have multiple properties, and thus a single physical structure (or set of physical structures) may be described and/or illustrated herein as multiple distinct circuit components.
  • a single piece of metal in a particular physical incarnation of an example CAM may serve as both a gate electrode of a transistor and as a wiring line.
  • two or more circuit components may be referred to or illustrated herein as distinct components should not be interpreted to mean that their corresponding physical structures in a physical incarnation of the example CAM are distinct structures.
  • first circuit component being “connected to” a second circuit component
  • a reference to a first circuit component being “connected to” a second circuit component could encompass: (A) a scenario in which a physical structure that serves as a terminal of the first circuit component is in direct physical contact with a physical structure that serves as a terminal of the second circuit; (B) a scenario in which a physical structure that serves as a terminal of the first circuit component is in direct physical contact with an electrical conductor (e.g., a wiring line) that is itself in direct physical contact with a physical structure that serves as a terminal of the second circuit; (C) a scenario in which the same physical structure that serves as a terminal of the first circuit component also serves as a terminal of the second circuit component; etc.
  • an electrical conductor e.g., a wiring line
  • first component being “connected between” second and third component
  • second component this means that two opposing terminals of the first component are connected to the second component and to the third component, respectively.
  • transistor being “connected between” two elements
  • applying a given voltage difference “across” a circuit component means applying voltages to two terminals of the circuit component such that the larger of the applied voltages minus the smaller of the applied voltages equals the given voltage difference.
  • a polarity with which the voltage difference is applied across the circuit component corresponds to the direction in which current would tend to flow through the circuit component (if allowed to do so) as a result of the applied voltage difference, which will depend upon which terminal of the circuit component has the larger of the applied voltages.
  • the given voltage difference when the given voltage difference is applied across the circuit component by applying the larger of the applied voltages to a first terminal of the circuit component and the smaller of the applied voltages to a second terminal of the circuit component, this may be referred to as applying the given voltage difference with a first polarity; when the given voltage difference is applied across the circuit component by applying the larger of the applied voltages to the second terminal and the smaller of the applied voltages to a first terminal, this may be referred to as applying the given voltage difference with a second polarity.
  • references herein to a resistance R LRS that is associated with a low resistance state of the memristor M should not be interpreted to imply that each memristor M is set precisely to the value R LRS each time they are set to the low resistance state.
  • the low resistance state of the memristor M might encompass a range of resistance values rather than just one specific resistance value.
  • the resistance of the memristor M is not necessarily the exact same value each time that the memristor M is set to the low resistance state. This may occur, for example, as a result of inevitable manufacturing variances between memristors M as well as inventible variances in the voltages that are applied for setting/resetting operations.
  • the resistance R LRS is a particular resistance value that is associated with the range of resistance values that comprises the low resistance state of the memristor M.
  • references herein to a resistance R HRS that is associated with a high resistance state of the memristor M should not be interpreted to imply that each memristor M is set precisely to the value R HRS each time they are reset to the high resistance state.
  • the high resistance state of the memristor M might encompass a range of resistance values rather than just one specific resistance value.
  • the resistance of the memristor M is not necessarily the exact same value each time that the memristor M is reset to the high resistance state.
  • the resistance R HRS is a particular resistance value that is associated with the range of resistance values that comprises the high resistance state of the memristor M.
  • the resistances R LRS and R HRS may be determined by a variety of methods. For example, a range of resistance values for the low resistance state and a range of resistance values for the high resistance state may be determined by repeatedly setting and resetting the memristor M and measuring its resistance each time and determining the ranges of values based on these measurements, and then the resistances R LRS and R HRS may be determined to be particular values within their corresponding ranges. For example, the resistance R LRS may be the highest value of the range of low resistance state values and the resistance R HRS may be the lowest value of the range of high resistance state values.
  • the resistances R LRS and R HRS may be determined to be statistical aggregations of measured resistance values of the memristor M (or of multiple memristors M) in the set and reset states, respectively; for example, R LRS and R HRS may be determined to be the average, the median, the mode, etc., of their respectively corresponding measured resistance values.
  • the memristor(s) M may be repeatedly set and reset and their resistances measured, and R LRS and R HRS may be determined to be the lowest and the highest resistance values, respectively, that are measured in this process.
  • the transistors of the bit cell were described as all being NFET transistors, but it should be understood that all PFET transistors could be used instead.
  • the voltages that result in the transistors turning on and off would be reversed (low gate-source voltage turning the PFET on and high gate-source voltage turning the PFET off), and therefore voltages applied during the search operations may be reversed to preserve the desired search results.
  • references to “a number” mean any integer greater than or equal to one.
  • the word describing the item(s) may be written with the pluralized “s” for grammatical consistency, but this does not necessarily mean that multiple items are being referred to.
  • a number of comparators could encompass both one comparator and multiple comparators.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Materials Engineering (AREA)
  • Computer Hardware Design (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Read Only Memory (AREA)

Abstract

An example content addressable memory. A bit cell of the memory may include a memristor and a switching transistor that are connected in series between a first data line and a second data line. The bit cell may also include a match-line transistor connected between a match line and a rail. A gate of the match-line transistor may be connected to a common node of the memristor and the switching transistor. The switching transistor may be sized such that its channel resistance when on is between a resistance associated with a low-resistance state of the memristor and a resistance associated with a high-resistance state of the memristor.

Description

BACKGROUND
Content addressable memory (CAM) is a type of memory that can perform a search operation in which a data string may be input as search content and the resulting output is an address of a location in the memory that stores matching data (if there is any). This is in contrast to a read operation in which an address is input and the resulting output is the data stored in the memory location corresponding to the searched address. Certain CAMs may be able to perform both the aforementioned search operation and the aforementioned read operation, while non-CAM memories may be able to perform the read operation but not the search operation.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1A is circuit diagram illustrating an example bit cell of an example CAM.
FIG. 1B is a circuit diagram illustrating an example resistive divider.
FIG. 2 is a table illustrating an association between resistance states and stored values.
FIG. 3 is a table illustrating search results for combinations of search criteria and stored value.
FIG. 4 is a table illustrating voltages that are applied to wiring lines of an example CAM to perform writing and reading operations.
FIG. 5A is a table illustrating voltages that are applied to wiring lines of an example CAM to perform search operations.
FIG. 5B is a signal diagram illustrating voltages that are applied to wiring lines of an example CAM to perform search operations, as well as the states of an example bit cell resulting from such search operations for each combination of search criterion and stored value.
FIG. 6 is a diagram illustrating an example CAM with an array of multiple bit cells.
FIG. 7 is a diagram illustrating an example CAM with an array of bit cells storing particular values.
FIG. 8 is a diagram illustrating an example device that includes an example CAM as a component thereof.
DETAILED DESCRIPTION
Described herein are example CAMs that include memristors as storage elements. Also described herein are example devices that may have a CAM as a component thereof, as well as example methods of operating CAMs.
[Examples CAMs—Overview]
The example CAMs described herein may include memristors as storage elements. The use of memristors as storage elements, as opposed to using, for example, SRAMs or DRAMs, allows the example CAMs disclosed herein to be non-volatile memory devices. In addition, as a result of using memristor-based storage elements, the example CAMs disclosed herein may have higher storage densities and lower power consumption than comparable CAMs that use SRAMs or DRAMs as storage elements.
Moreover, certain of the example CAMs disclosed herein may be capable of being searched on the basis of three search criteria (e.g., 0, 1, and a wildcard), which may provide increased flexibility and speed in search operations as compared to a CAM that is searchable on the basis of two search criteria (e.g., 0 and 1).
Furthermore, the example CAMs disclosed herein may have bit cell circuit layouts that can fit in a smaller layout space than other CAM bit cell layouts, which may allow the example CAMs to have a more densely packed memory array (i.e., more bit cells per unit area) than other comparable CAMs. In certain examples, such savings of bit cell layout space might be achieved, for example, as a result of the bit cell circuit layout allowing for the use of just two FETs per bit cell instead of the three or more FETs per bit cell that may be required in other circuit layouts in order to provide CAM functionality. As another example of how such savings of bit cell layout space might be achieved, in certain of the example CAMs disclosed herein all of the transistors in the bit cells of the CAM may be a same channel type as one another (for example, all NFETs); this may allow for a well to be omitted from the bit cell array that might otherwise be needed, thus freeing up layout space that would have been used for the well. In addition to saving layout space, the use of transistors having all the same channel type may also decrease the complexity and/or cost of manufacturing the CAM.
In particular, in certain example CAMs disclosed herein, a bit cell may include a memristor, a switching transistor, and a match-line transistor. The memristor and the switching transistor may be connected in series between a first data line and a second data line. The match-line transistor may be connected between a match-line and a rail. A gate of the match-line transistor may be connected to a common node of the memristor and the switching transistor. The switching transistor may be sized such that its channel resistance when on (i.e., when ohmic) is between a resistance associated with a low-resistance state of the memristor and a resistance associated with a high-resistance state of the memristor. A gate of the switching transistor may be connected to a word line.
Furthermore, certain of the example CAMs may also include control circuitry that searches the bit cell based on a first search criterion (e.g., 0), based on a second search criterion (e.g., 1), and based on a wildcard search criterion. If the bit cell stores a value that does not match the search criterion, the voltage of the common node during the search may become a voltage that turns on the match-line transistor, thereby connecting the match line to the rail and resulting in a voltage of the match line (which has been pre-charged) being pulled down below its pre-charged voltage. If the bit cell stores a value matching the search criterion, then the voltage of the common node may become a voltage that keeps the match-line transistor off so that the voltage of the match-line is not pulled down (at least not by that bit cell—it might be pulled down by some other bit cell). Thus, matches or misses of the search criterion may be determined by monitoring the match line and determining whether or not its voltage is pulled low.
The control circuit may search the bit cell based on the first search criterion by pre-charging the match line, applying a first voltage to the first data line, applying a ground voltage to the second data line, and applying a third voltage to the rail. The control circuit may search the bit cell based on a second search criterion by pre-charging the match line, applying the ground voltage to the first data line, applying the first voltage to the second data line, and applying a fourth voltage to the rail. The control circuit may search the bit cell based on the wildcard search criterion by applying a ground voltage to both the first and second data lines.
The memristor and the switching transistor of the bit cell may form a resistive divider. An output voltage of the resistive divider (i.e., the voltage of the common node between the memristor and the switching transistor) is applied to the gate of the match-line transistor, and thus the output voltage of the resistive divider may control whether the match-line transistor is on or off. The relative resistances of the memristor and the switching transistor and the voltages that are applied to the bit cell during a search operation may be set such that the output voltage of the resistive divider during the search operation is an off-voltage (low voltage) when the memristor stores a value that matches the search criterion and an on-voltage (high voltage) when the memristor stores a value that matches the search criterion. When the search criterion is the wildcard search criterion, the output voltage of the resistive divider is low regardless of what value is stored in the resistive divider.
Certain example CAMs may include multiple instances of the example bit cells described above, which may be arrayed in rows and columns. In certain examples, each first and second data line may be connected to each bit cell in a same column, and each word line may be connected to each bit cell in a same row. Each match line may be connected to multiple bit cells that are in a same row as one another, with the number of bit cells per match line being equal to a search unit size for the CAM (i.e., a word size) and the bit cells that are connected to the same match line as one another forming a word storage block. The bit cells of the word storage block may be searched to see if the data stored in the word bit cells of the word storage block matches an input search word. The match line for a given word storage block may be pulled low if any one of the bit cells thereof misses its respective search criterion, and therefore it can be determined whether the given word storage block stores data that matches the search word by monitoring whether the corresponding match line is pulled low. Each row may include a single word storage block (in which case there may be one word line per row), or each row may include P>1 word storage blocks (in which case there may be P match lines per row).
[Bit Cells—Detailed Examples]
FIG. 1A illustrates an example bit cell 110 of an example CAM. The example bit cell 110 may include a memristor M, a transistor T1, and a match-line transistor T2. The memristor M may be connected between a first data line DL1 and a node G. The transistor T1 may be connected between the node G and a second data line DL2. Thus, the memristor M and the transistor T1 may be connected via a common node, namely the node G. The match-line transistor T2 may be connected between a rail RL and a match line ML. A gate of the match-line transistor T2 may be connected to the node G, and therefore a voltage of the node G may control a conductive state of the match-line transistor T2. A gate of the transistor T1 may be connected to a word line WL, and therefore a voltage of the word line WL may control a conductive state of the transistor T1.
The memristor M may serve as the storage element of the bit cell 110. The memristor M may be a device whose resistance can be changed between multiple resistance states by applying certain voltage differences across the memristor M or flowing certain currents through the memristor M, with the memristor M “remembering” or maintaining its most recent resistance state even when the voltage/current that caused it to enter that state has been removed. In other words, the current resistance state of the memristor M may depend on the magnitude and polarity of voltage differences that have been applied across the memristor in the past or on how much current has flowed in what direction through the memristor M in the past.
For example, the memristor M may be capable of changing between a low resistance state in which the memristor M exhibits relatively low resistance and a high resistance state in which the memristor M exhibits relatively high resistance. The memristor M may be caused to change to the low resistance state by applying a setting voltage difference of sufficient magnitude with a first polarity across the memristor, and the memristor M may be caused to change to the high resistance state by applying a resetting voltage difference of sufficient magnitude across the memristor M with a second polarity.
Once the memristor M is set or reset, it will tend to stay in its current resistance state until it is reset or set again. As a result, the resistance state of the memristor M may be used to store data values. For example, a high resistance state of the memristor M may represent a first data value and a low resistance state of the memristor M may represent a second data value. Herein, it is assumed that a high resistance state represents a digital 0 and a low resistance state represents a digital 1, as illustrated in the table of FIG. 2; however, it should be understood that the logical association between resistance state and stored value may be arbitrarily selected.
The memristor M may be set to a low resistance state by turning on the switching transistor T1 (e.g., apply VDD to the word line WL), applying the setting voltage Vset to the first data line DL1, and applying a ground voltage GND to the second data line DL2, as illustrated in FIG. 4. The memristor M may be reset to a high resistance state by turning on the switching transistor T1 (e.g., apply VDD to the word line WL), applying the resetting voltage Vreset to the second data line DL2, and applying the ground voltage GND to the first data line DL1, as illustrated in FIG. 4. The magnitudes of Vset and Vreset may be set to any values that are larger than minimum values required to be able to set and reset the memristor M, respectively. In particular, if Vset _ min is the minimum voltage that will result in setting the memristor M and Vreset _ min is the minimum voltage that will result in resetting the memristor M, then Vset may be equal to or greater than Vset _ min and Vreset may be equal to or greater than Vreset _ min. Vset and Vreset do not necessarily need to have the same magnitude, since Vset _ min is not necessarily the same as Vreset _ min. However, Vset and Vreset could be set to the same value as long as that value is greater than or equal to both Vset _ min and Vreset _ min. In certain examples, VDD=Vset=Vreset, where VDD is a highest voltage level of the TCAM. The values of Vset _ min and Vreset _ min may be parameters of the memristor M that are known in advance, that are determined theoretically based on a model of the memristor M, or that are determined experimentally.
The value that is stored in the bit cell 110 may be read by applying a read voltage difference across the memristor M and measuring a current that flows through the memristor M. The magnitude of the current that flows through the memristor M may be used to determine the resistance state of the memristor M because a high resistance state will result in low current, while a low resistance state will result in high current. In particular, the value stored in the bit cell may be read by turning on the switching transistor T1 (e.g., apply VDD to the word line WL), applying the read voltage Vread to either one of the first and second data lines DL1/DL2, and applying the ground voltage GND to the other one of the first and second data lines DL1/DL2, as illustrated in FIG. 4. The current that flows through the memristor M may be sensed on either one of the data lines DL1/DL2, or the current flowing on the match line ML may be sensed. In general, applying voltage differences across the memristor M whose magnitudes are relatively small in comparison to the setting/resetting voltage differences will not significantly affect the resistance state of the memristor M. Thus, in order to ensure that reading the value stored in the bit cell 110 does not change a resistance state of the memristor M, the read voltage difference may be set to a value that is below a minimum voltage that will set or reset the memristor M.
Instead of measuring the current via first and second data lines DL1/DL2 to determine a resistance state of the memristor M, the resistance state of the memristor M may be determined by measuring a voltage of the match line ML during a reading operation. The match-line transistor T2 may be set to act as an amplifier (e.g., common source amplifier or source-follower amplifier), and the voltage of the match line ML may vary depending on the voltage of the node G. Since the voltage of the node G during a reading operation depends on the resistance state of the memristor, the voltage of the match line ML will also depend on the resistance state of the memristor.
The bit cell 110 may be thought of as including a resistive divider 111, with the memristor M and the switching transistor T1 serving as the resistors of the resistive divider 111. In general, a resistive divider is formed when two resistors are connected in series, input voltages are applied to opposite input terminals of the resistors, and an output voltage is taken from a common node between the resistors. In particular, FIG. 1B illustrates a generalized form of a resistive divider, with resistors R1 and R2 being connected in series, input voltages VIN _ 1 and VIN _ 2 being applied to the input terminals, and the output voltage VOUT being taken from the common node between the resistors R1 and R2 (which may be referred to as the output terminal). Using the labels from FIG. 1B, the output voltage of a generalized resistive divider may be given by the following:
V OUT = R 2 V IN _ 1 + R 1 R IN _ 2 R 1 + R 2 ( eq . 1 )
The memristor M and the switching transistor T1 may be considered as being the resistors of the resistor divider 111, and correspond, respectively, to the resistors R1 and R2 of the generalized resistive divider. In a certain sense, a memristor is not the same type of circuit component as a resistor, as the memristor can have its resistance changed over time while a resistor generally has a constant resistance. However, during periods of time in which voltage differences that are less than the setting or resetting voltage differences are applied across a memristor, the memristor has an approximately constant resistance, and thus the memristor M may be modeled approximately during these periods as a resistor. The switching transistor T1 may be sized such that it has a non-negligible channel resistance when the transistor T1 is on (i.e., when it is ohmic), and this channel resistance may be modeled as a resistor.
In particular, the switching transistor T1 may be sized such that its channel resistance when on (RT1) is greater than a resistance associated with a low resistance state of the memristor M (RLRS), but less than a resistance associated with a high resistance state of the memristor M (RHRS)—that is, RLRS<RT1<RHRS. The particular value that is set for the channel resistance RT1 may be determined based on considerations described in greater detail below with regard to searching operations.
In the resistive divider 111, the two input terminals thereof are formed from the terminal of the memristor M that is connected to the first data line DL1 and the terminal of the transistor T1 that is connected to the second data line DL2. Thus, the input voltages of the resistive divider 111 are the voltage carried by the first data line DL1 (VDL1) and the voltage carried by the second data line DL2 (VDL2). The input voltages VDL1 and VDL2 of the resistive divider 111 correspond, respectively, to the input voltage VIN _ 1 and VIN _ 2 of the generalized resistive divider. The output terminal of the resistive divider 111 is the node G, and the output voltage of the resistive divider 111 is the voltage of the node G (VG), which corresponds to the output voltage VOUT of the generalized resistive divider. Thus, the voltage of the node G when the transistor T1 has been turned on may be given by the following (obtained from substituting corresponding variables into equation 1):
V G = R T 1 V D L 1 + R M R D L 2 R M + R T 1 ( eq . 2 )
where RM is the present resistance of the memristor M.
The example bit cell 110 may be searched based on three search criteria—namely 0, 1, and a wildcard (denoted Y herein for convenience). The configuration of the bit cell 110 and the voltages that are applied to the bit cell during a search operation are set such that search operations produce the search results illustrated by the table in FIG. 3. In particular, when the bit cell 110 is searched based on the 0 search criterion, the result is a match if the bit cell 110 stores a 0 (i.e., if RM≈RHRS) and a miss if the bit cell 110 stores a 1 (i.e., if RM≈RLRS). When the bit cell 110 is searched based on the 1 search criterion, the result is a miss if the bit cell 110 stores a 0 (i.e., if RM≈RHRS) and a match if the bit cell 110 stores a 1 (i.e., if RM≈RLRS). When the bit cell 110 is searched based on the wildcard search criterion (Y), the result is a match regardless of what value is stored by the bit cell 110.
FIGS. 5A and 5B illustrate how search operations based on the above-noted search criteria may be performed, as well as states of the bit cell in response to such search operations for each combination of search criteria and stored value.
To search the bit cell 110 based on the 0 search criterion, the match line ML may be pre-charged to a pre-charge voltage Vpre, the first transistor T1 may be turned on by applying VDD to the word line WL, a search voltage Vsearch may be applied to the first data line DL1 while the ground voltage GND is applied to the second data line DL2, and a voltage V0 may be applied to the rail RL, as illustrated in FIGS. 5A and 5B. During this search, the input voltages of the resistive divider 111 are VDL1=Vsearch and VDL2=GND. Thus, from equation 2 the voltage of the node G during a search based on the 0 search criterion (denoted by VG|S=0) becomes:
V G | S = 0 = R T 1 R M + R T 1 V search ( eq . 3 )
To search the bit cell 110 based on the 1 search criterion, the match line ML may be pre-charged to a pre-charge voltage Vpre, the first transistor T1 may be turned on by applying VDD to the word line WL, the ground voltage GND may be applied to the first data line DL1 while the search voltage Vsearch is applied to the second data line DL2, and a voltage V1 may be applied to the rail RL. During this search, the input voltages of the resistive divider 111 are VDL1=GND and VDL2=Vsearch. Thus, from equation 2 the voltage of the node G during a search based on the 1 search criterion (denoted by VG|S=0) becomes:
V G | S = 1 = R M R M + R T 1 V search ( eq . 4 )
The voltage of the node G may control a conductive state of the match-line transistor T2—in particular, the match-line transistor T2 may be turned on when a gate-source voltage thereof is greater than a threshold voltage thereof (Vth), and may be kept off when a gate-source voltage thereof is less than Vth. As can be seen from equations 3 and 4, the voltage of the node G during a search operation will depend on the resistance state of the memristor M. Thus, the conductive state of the match-line transistor T2 during a search operation may be controlled based on the resistance state of the memristor M. In particular, by appropriately setting the various resistances of the bit cell 110 and the voltages applied thereto during the search operation, it can be assured that the search results illustrated in FIG. 3 are obtained.
In particular, if the search criterion is 0 and the bit cell 110 stores a 0, then, using equation 3 and RM≈RHRS, the voltage of the node G in this state (denoted by
V G | D = 0 S = 0 )
will become:
V G | D = 0 S = 0 R T 1 R HRS + R T 1 V search ( eq . 5 )
This state is illustrated in FIG. 5B in the column labeled “search criterion=0” and the row labeled “stored bit=0”. As illustrated in FIG. 5B,
V G | D = 0 S = 0
corresponds to a relatively low voltage, since RT1<RHRS. In particular,
V G | D = 0 S = 0 < 1 2 V search .
Thus, if V0 is appropriately set, the gate-source voltage of the match-line transistor T2 may be kept below the threshold voltage Vth, and therefore the match-line transistor T2 may be kept off. As a result, the bit cell 110 does not pull the voltage of the match line ML down, thereby indicating a match.
If the search criterion is 0 and the bit cell 110 stores a 1, then, using equation 3 and RM≈RLRS, the voltage of the node G in this state (denoted by
V G | D = 1 S = 0 )
will become:
V G | D = 0 S = 0 R T 1 R LRS + R T 1 V search ( eq . 6 )
This state is illustrated in FIG. 5B in the column labeled “search criterion=0” and the row labeled “stored bit=1”. As illustrated in FIG. 5B,
V G | D = 1 S = 0
corresponds to a relatively high voltage, since RLRS<RT1. In particular,
1 2 V search < V G | D = 1 S = 0 .
Thus, if V0 is appropriately set, the gate-source voltage of the match-line transistor T2 in this state may be raised above the threshold voltage Vth, and the match-line transistor T2 may be turned on. As a result, the match line ML is connected to the rail RL and begins to discharge, thus pulling the voltage of the match line ML low (thereby indicating a miss).
If the search criterion is 1 and the bit cell 110 stores a 0, then, using equation 4 and RM≈RHRS, the voltage of the node G in this state (denoted by
V G | D = 0 S = 1 )
becomes:
V G | D = 0 S = 1 R HRS R HRS + R T 1 V search ( eq . 7 )
This state is illustrated in FIG. 5B in the column labeled “search criterion=1” and the row labeled “stored bit=0”. As illustrated in FIG. 5B,
V G | D = 0 S = 1
corresponds to a relatively high voltage, since RT1<RHRS. In particular,
1 2 V search < V G | D = 0 S = 1 .
Thus, if V1 is appropriately set, the gate-source voltage of the match-line transistor T2 in this state may be raised above the threshold voltage Vth, and the match-line transistor T2 may be turned on. As a result, the match line ML is connected to the rail RL and begins to discharge, thus pulling the voltage of the match line ML low (thereby indicating a miss).
If the search criterion is 1 and the bit cell 110 stores a 1, then, using equation 4 and RM≈RLRS, the voltage of the node G in this state (denoted by
V G | D = 1 S = 1 )
becomes:
V G | D = 1 S = 1 R LRS R LRS + R T 1 V search ( eq . 8 )
This state is illustrated in FIG. 5B in the column labeled “search criterion=1” and the row labeled “stored bit=1”. As illustrated in FIG. 5B
V G | D = 1 S = 1
corresponds to a relatively low voltage, since RHRS<RT1. In particular,
V G | D = 1 S = 1 < 1 2 V search .
Thus, if V0 is appropriately set, the gate-source voltage of the match-line transistor T2 may be kept below the threshold voltage Vth, and therefore the match-line transistor T2 may be kept off. As a result, the bit cell 110 does not pull the voltage of the match line ML down, thereby indicating a match.
In order to ensure that misses result in the match-line transistor T2 being turned on and matches result in the match-line transistor T2 being kept off, then the voltages V0, V1, and Vsearch, the resistances RT1, RHRS, RLRS, and the threshold voltage Vth may be set such that:
R T 1 R HRS + R T 1 V search - V th < V 0 < R T 1 R LRS + R T 1 V search - V th ; and R LRS R LRS + R T 1 V search - V th < V 1 < R HRS R HRS + R T 1 V search - V th ( eq . 9 )
In certain examples, the resistances RT1, RHRS, RLRS and/or the threshold voltage Vth may be set based on other considerations, in which case the conditions of equation 9 may be satisfied by setting V0, V1, and Vsearch. In certain examples, V0=V1=½ Vsearch−Vth.
In certain examples, it may be desirable for Vsearch to be less than Vset _ min and less than Vreset _ min, so that the search operation does not inadvertently set or reset the memristor M. For example, Vsearch may be the same value as Vread, since Vread is less than Vset _ min and Vreset _ min. Using the same voltage for Vread and Vsearch may reduce the complexity of drivers that are used to supply the voltages to the bit cell 110, as fewer distinct voltages may need to be generated.
To search the bit cell 110 based on the wildcard (Y) search criterion, the match line ML may be pre-charged to a pre-charge voltage Vpre, the first transistor T1 may be turned on by applying VDD to the word line WL, and the ground voltage GND may be applied to the first data line DL1, the second data line DL2, and the rail RL. During this search, the input voltages of the resistive divider 111 are VDL1=GND and VDL2=GND. Thus, the voltage of the node G during a search based on the Y search criterion becomes GND, regardless of what value is stored in the bit cell 110. Because the voltage of the node G during a Y search is always GND and because GND is applied to the rail RL. the match-line transistor T2 is kept off (indicating a match) regardless of the value stored in the bit cell 110.
As illustrated in FIG. 5B, during a searching operation it may be desirable to turn on the transistor T1 while the ground voltage GND is applied to the data lines DL1/DL2 and before the search voltage Vsearch is applied to either data line DL1/DL2. This is to avoid having Vsearch temporarily appear on the node G before VG can stabilize to its appropriate value, thereby mistakenly turning on the match-line transistor T2 temporarily and possibly discharging the match line ML. By turning on the transistor T1 while GND is applied to both data lines DL1/DL2, it is ensured that GND is applied to the node G at the start of the search operation.
[CAMs—Detailed Examples]
FIG. 6 illustrates an example CAM 100 that includes an array 140 of multiple bit cells 110. The bit cells 110 may be arrayed in W rows and N columns. In FIG. 6, each row corresponds to a word storage block (e.g., word 0 through word W−1), with each bit cell 110 in a given row corresponding to a bit position (e.g., bit 0 through bit N−1). The CAM 100 may include multiple word lines WL (e.g., WL0 though WLW-1), multiple match lines ML (e.g., ML0 though MLW-1), multiple first data lines DL1 (e.g., DL1 0 though DL1 N-1), multiple second data lines DL2 (e.g., DL2 0 though DL2 N-1), and multiple rails RL (e.g., RL0 through RLN-1). Each bit cell 110 in a given row may be connected to the same word line WL and to the same match line ML as one another. Thus, all of the bit cells 110 in a given word may be selected at the same time for writing/reading by turning on the word line WL that corresponds to the given word. In addition, during a search operation the match line ML of a given word may be pulled low if any one of the bit cells 110 that correspond to the word is a miss. Each bit cell 110 in a given column may be connected to the same first data line DL1, the same second data line DL2, and the same rail RL as one another. Thus, each bit cell 110 in a same column may be searched simultaneously based on the same search criterion.
The CAM 100 may include control circuitry 150 that applies signals to the various lines mentioned above so as to control operations of the bit cells, such as read, write, and search operations. The control circuitry 150 may comprise drivers, shift registers, decoders, and/or the like that are connected to the various wiring lines of the array 140. The control circuitry 150 may receive a control signal Ctrl that instructs the control circuitry 150 which operation it should perform. The control circuitry 150 may also receive a read/write address that indicates a target word storage block for a read operation or a write operation. The control circuitry 150 may output an output word as a result of performing a read operation, with the output word corresponding to the word that is stored in the word storage block whose address corresponds to the read/write address. The control circuitry 150 may also receive an input word that indicates either a word that the CAM 100 is to be searched for during a search operation or a word that is to be written to a particular word storage block during a write operation. The control circuitry 150 may also output an output address as a result of performing a search operation, with the output address corresponding to the address (or addresses) of any word storage block(s) whose stored word matches the input search word.
In FIG. 6, the control circuitry 150 is illustrated as including a row driver 151 and a column driver 152. The row driver 151 is connected to the word lines WL0 though WLW-1 and the match lines ML0 though MLW-1, and may supply electrical signals to and/or sense electrical signals carried on these wiring lines. The column driver 152 is connected to the first data lines DL1 0 though DL1 N-1, the second data lines DL2 0 though DL2 N-1, and the rails RL0 through RLN-1-1, and may supply electrical signals to and/or sense electrical signals carried on these wiring lines. It should be understood that this representation of the control circuitry 150 is merely for ease of description, and that it does not imply that the control circuitry 150 is necessarily composed of exactly two drivers in the configuration illustrated. For example, the control circuitry 150 could include a separate driver for each type of wiring line, a single driver for all the wiring lines, or any combination of drivers. As another example, the illustrated locations of the drivers relative to the array 140 are not intended to be meaningful, and it should be understood that drivers could be located on any side or combination of sides of the array 140.
The control circuitry 150 may control the CAM 100 to perform a read operation and a write operation in units of words. In particular, when the control signal Ctrl instructs the control circuitry 150 that a read operation or a write operation is to be performed, the row driver 151 may select the target word storage block based on the input read/write address by applying VDD to its corresponding word line WL, and the column driver 152 may then proceed to apply voltages to the column lines to perform the read/write operations for each bit cell 110 in the selected word storage block.
In particular, during a read operation, the column driver 152 may apply voltages to each of the first and second data lines DL1/DL2 and rails RL according to the table illustrated in FIG. 4. The current flowing through each of the N bit cells 110 of the target word storage block may be sensed by the column driver 152 (for example, via either the first or second data lines DL1/DL2). The column driver 152 may determine a bit value stored in each bit cell 110 based on the current that flows therethrough, and these N bit values may be combined into the output word that is output from the column driver 152.
During a write operation, the column driver 152 may determine, based on the input word, whether each of the N bit cells 110 in the target word storage block is to be set or reset and, in accordance with this determination, set or reset each of the N bit cells 110 in the target word storage block by applying the appropriate voltages to the corresponding first and second data lines DL1/DL2 and rail RL. In certain examples, setting and resetting operations may be performed simultaneously across the target word storage block. In certain other examples, setting and resetting operations may be performed in distinct phases. For example, all of the bit cells 110 in the target word storage block may first be reset, and then those bit cells 110 that are to be set may subsequently be set.
Although the reading and writing operations were described above as being performed in units of words, it should be understood that writing could be performed in other units. For example, any number of bit cells 110 could be written to or read from during a read or write operation without necessarily writing to or reading the other bit cells 110 in the same word storage block. In such an example, the word line WL of the word storage block that contains the target bit cell(s) 110 could be selected, the ground voltage GND could be applied to all of the first and second data lines DL1/DL2 that correspond to bit cells 110 that are not the target bit cell(s), and the voltages for the writing/reading operation could be applied to the first and second data lines DL1/DL2 of the target bit cell(s) 110. Thus, even though an entire word storage block is selected by its corresponding word line WL, any bit cells 110 that are not desired to be writing to or read from can be omitted from the writing/reading operation simply by applying GND to their corresponding data lines DL1/DL2.
The control circuitry 150 may control the CAM 100 to perform a search operation across the entire array 140 simultaneously. In particular, when the control signal Ctrl instructs the control circuitry 150 that a search operation is to be performed, the row driver 151 may select all of the target word storage blocks simultaneously by applying VDD to all of the word lines WL, and may pre-charge all of the match lines ML. The column driver 152 may then proceed to apply voltages to the column lines to perform the search operations for each bit position based on the input search word. In particular, the column driver 152 may determine, based on the input search word, which search criterion is to be used to search each bit position, and in accordance with this determination, apply the appropriate voltages to the corresponding first and second data lines DL1/DL2 and rail RL. The row driver 151 may then monitor the match lines to determine which are pulled low (indicating the corresponding word storage block stores a word that is a miss) and which remain charged (indicating the corresponding word storage block stores a word that is a match). The row driver 151 may decode the results of misses/matches from the match lines ML and output an address (or addresses) of a matching word storage block (or matching words storage blocks).
For example, consider the example CAM 100 illustrated in FIG. 7, which includes an array 140 having three rows and three columns of bit cells 110, with the bit cells 110 each storing values as indicated in the figure. Suppose that this example CAM 100 is searched with the input search word being “Y10”. The column driver 152 will receive the input search word and the control signal Ctrl indicating a search operation, and based thereon would determine that the bit-0 column should be searched based on the wildcard search criterion (since the 0th bit of the search word is Y), the bit-1 column should be searched based on the 1 search criterion (since the 1st bit of the search word is 1), and the bit-2 column should be searched based on the 0 search criterion (since the 2nd bit of the search word is 0). Accordingly, the column driver 152 may perform the search operation by applying voltages to the various wiring lines as follows:
bit-0 column bit-1 column bit-2 column
(Y search criterion) (1 search criterion) (0 search criterion)
Line: DL10 DL20 RL0 DL11 DL21 RL1 DL12 DL22 RL2
Voltage: GND GND GND GND Vsearch V1 Vsearch GND V0
Since the bit-0 column is searched on the basis of the wildcard criterion, none of the bit cells 110 in this column will be a miss, and hence none of these bit cells 110 will pull down their respective match lines ML. In particular, the voltage of the node G in each bit cell 110 in the bit-0 column will be GND because GND is applied to both DL1 0 and DL2 0. Because GND is also applied to RL0, the gate-source voltage of the match-line transistor T2 of each bit cell 110 in the bit-0 column will be zero, and hence the match-line transistor T2 of each bit cell 110 in the bit-0 column will remain off.
The [bit-1, word-0] bit cell 110 stores a 1 and is searched on the basis of the 1 search criterion, and hence from equation 8 the voltage of node G becomes
R LRS R LRS + R T 1 V search ,
which is a low value. Thus, the match-line transistor T2 remains off and the [bit-1, word-0] bit cell 110 does not pull down its match line ML0.
The [bit-2, word-0] bit cell 110 stores a 0, and is searched on the basis of the 0 search criterion, and hence from equation 5 the voltage of node G becomes
R T 1 R HRS + R T 1 V search ,
which is a low value. Thus, the match-line transistor T2 remains off and the [bit-2, word-0] bit cell 110 does not pull down its match line ML0.
Because none of the bit cells 110 in the word 0 pull down the match line ML0, the row driver 151 determines that the word 0 is a match for the search word. Thus, an address of the word 0 is output from the row driver 151 as an output address.
The [bit-1, word-1] bit cell 110 stores a 0 and is searched on the basis of the 1 search criterion, and hence from equation 6 the voltage of node G becomes
R HRS R HRS + R T 1 V search
which is a high value. Thus, the match-line transistor T2 is turned on and the [bit-1, word-1] bit cell 110 pulls down its match line ML1.
The [bit-2, word-1] bit cell 110 stores a 1 and is searched on the basis of the 0 search criterion, and hence from equation 7 the voltage of the node G becomes
R T 1 R LRS + R T 1 V search ,
which is a high value. Thus, the match-line transistor T2 is turned on and the [bit-2, word-1] bit cell 110 pulls down its match line ML1.
Because the [bit-1, word-1] and [bit-2, word-1] bit cells 110 pull down the match line ML1, the row driver 151 determines that the word 1 is a miss for the search word. Thus, an address of the word 1 is not output from the row driver 151 as an output address.
The states of the [bit-1, word-2] and [bit-2, word-2] are similar to the states of the [bit-1, word-0] and [bit-2, word-0] bit cells 110 that were described above, and hence they will not be described in detail. In short, since the [bit-1, word-2] and [bit-2, word-2] bit cells 110 are matches, neither one pulls down the match line ML2, and therefore the row driver 151 determines that the word 2 is a match for the search word. Thus, an address of the word 2 is output from the row driver 151 as an output address.
[Example Devices with CAMs]
The example CAMs described herein may be used in any electronic device in which a memory might be useful, such as, for example, in a personal computer, server, smartphone, tablet, network device, etc. In particular, the example CAMs can be used to store data for the device by writing the data into the CAM, and the device can access the stored data by reading from the CAM. Moreover, the example CAMs may also allow the device to search the CAM based on an input search word, which can be particularly useful in a variety of applications. For example, a database application executed by an example device may benefit greatly by the device using the example CAMs. As another example, a network device (such as a router, switch, gateway, etc.) may greatly benefit from using the example CAMs, as the high search speed provided by the CAM may allow the network device to operate at higher speeds than may otherwise be possible.
FIG. 8 illustrates an example device 500 that includes an example CAM 100. For ease of explanation, certain aspects of the example device 500 will be described below with the assumption that the example device 500 operates as a network device, but it should be understood that the example device 500 could be any type of electronic device. Moreover, some of the features described below could be omitted from the example device 500 and others not described below could be added.
The device 500 may include a CAM 100, a processing circuitry 510, communications circuitry 520, and machine readable media 530. The CAM 100 may be an instance of the example CAM 100 described above, and may include an array 140 of bit cells 110 and control circuitry 150.
The processing circuitry 510 may supply the control signal Ctrl to the control circuitry 150. The processing circuitry 510 may also supply the input word and the read/write address to the control circuitry 150. The processing circuitry 510 may be any circuitry capable of executing machine-readable instructions, such as a central processing unit (CPU), a microprocessor, a microcontroller device, a digital signal processor (DSP), etc. The processing circuitry 510 may also be an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), an application-specific instruction set processor (ASIP), or the like, that is configured to perform certain operations described herein, such as the read, write, and search operations.
The machine readable media 530 may be any non-transitory machine readable medium, which may include volatile storage media (e.g., DRAM, SRAM, etc.) and/or non-volatile storage media (e.g., PROM, EPROM, EEPROM, NVRAM, hard drives, optical disks, etc.). The machine readable media 530 may store machine-readable instructions that, when executed by the processing circuitry 510, cause the device 500 to perform some or all of the operations described herein, such as the read, write, and search operations.
The communications circuitry 520 may be circuitry for receiving input data communications and sending output data communications. For example, the communications circuitry 520 may include a network interface card. In certain examples, the communications circuitry 520 may include multiple communications ports 521, and may serve to connect multiple other electronic devices to one another via the device 500. For example, the device 500 may be a network router, network switch, network gateway, or the like, and may perform various traffic control tasks such as routing, switching, etc. In certain examples, the communications circuitry 520 may receive communications that include a destination address (such as an IP address), and the communications circuitry 520 may determine which communications port 521 to forward the received communication to based on the destination address.
In particular, the communications circuitry 520 may determine which port 521 to forward a received communication to by searching the CAM 100 based on the destination address that was included in the communication. For example, each device that is connected to the device 500 may have a unique device address (such as an IP address), and the device 500 may store the device address of each device that is connected thereto in the CAM 100 such that the location of the stored device address within the CAM 100 corresponds to the communications port 521 to which the device is connected. In particular, each word storage block of the CAM 100 may correspond to a different port 521 of the device 500, and whenever a new device is connected to the device 500 the device address of the new device may be written into a word storage block that corresponds to the port 521 to which the new device is connected. Subsequently, when a communication is received by the communication circuitry 520, the destination address of the communication may be sent to the CAM 100 (either directly, or via the processing circuitry 510) as an input search word; the CAM 100 may then return a memory address of a matching word storage block to the communications circuitry 520 (either directly or via the processing circuitry 510). Because each word storage block in this example corresponds to a particular port 521, the memory address of the word storage block may be understood by the communications circuitry 520 to identity a particular port 521, and therefore the communications circuitry 520 can determine which port 521 to forward the communication to based on the output address of the CAM 100. In certain examples, the memory address of each word storage block may be the same as the address of one of the ports 521, which may enable the communications circuitry 520 to easily determine from an output memory address which port 521 corresponds thereto.
In such an example, the CAM 100 may operate as sort of look-up table that indicates which port corresponds to which destination address. However, because the entire array 140 of the CAM 100 may be searched simultaneously and may be searched at the hardware level, the amount of time it takes for the device 500 to determine which port to forward a communication to can be greatly reduced.
In certain examples described above, the example memristor M is a bipolar memristor—i.e., it is set by a voltage difference/current having one polarity and reset by a voltage difference/current having a different polarity. However, in other examples the memristor M may be a unipolar memristor—i.e., a memristor that is set and reset by the same polarity. In examples that use unipolar memristors, the same reading and searching operations that were described above with regard to a bipolar memristor could be used, and any known unipolar memristor writing method could be used in place of the writing method described above.
In the description above, the memory element of the bit cell 110 is described as a memristor. In the art, the term “memristor” may be used in certain contexts in a broad sense and may be used in certain contexts in a narrow sense. In one narrow sense, “memristor” may refer specifically to circuit elements that exhibit a non-linear relationship between electric charge and magnetic flux (or exhibit a relationship between the time integral of current and the time integral of voltage). In the broad sense, “memristor” refers broadly to any non-volatile memory element that is based on changing resistance states and is read by its resistance. Herein and in the appended claims, “memristor” is always used in the broad sense, unless specifically indicated otherwise; specifically, as used herein, “memristor” refers broadly to any non-volatile memory element that is based on changing resistance states and is read by its resistance. Thus, as used herein the memristor M may be, for example, a memristor in the narrow sense of the term, a memory element that changes the resistance across a dielectric solid-state material, a phase-change memory element (sometimes referred to as PCM or PCRAM), a conductive-bridging memory element (sometimes referred to as CBRAM or a programmable metallization cell), a magnetoresistive memory element (sometimes referred to as MRAM), or the like.
The example CAMs are described herein and illustrated in the drawings in a conceptual or schematic manner to aid understanding. In particular, physical structures in the example CAMs are referred to and/or illustrated conceptually herein as circuit components, and the relationships between these circuit components are illustrated in circuit diagrams in accordance with the usual practice in the art. Circuit components are conceptual representations of classes of physical structures or devices that perform certain functions and/or have certain properties. Examples of such circuit components include passive devices such as resistors, capacitors, memristors, etc.; active devices such as transistors, diodes, etc.; constituent elements of the active/passive devices such as terminals, electrodes, gates, sources, drains, etc.; elements that connect devices such as wiring lines, nodes, etc.; and so on. It should be understood that a single physical structure (or set of physical structures) in an actual physical incarnation of an example CAM may serve multiple functions and/or have multiple properties, and thus a single physical structure (or set of physical structures) may be described and/or illustrated herein as multiple distinct circuit components. For example, a single piece of metal in a particular physical incarnation of an example CAM may serve as both a gate electrode of a transistor and as a wiring line. Thus, the fact that two or more circuit components may be referred to or illustrated herein as distinct components should not be interpreted to mean that their corresponding physical structures in a physical incarnation of the example CAM are distinct structures.
When reference is made herein or in the appended claims to a first circuit component being “connected to” a second circuit component, this means that: (1) the physical structures corresponding to the first and second components are so arranged that a current path exists therebetween, and/or (2) a single physical structure serves as at least a part of both the first and second circuit components. Note that, in light of this definition, a reference herein to or illustration in the drawings of multiple circuit components being “connected to” one another does not imply that the circuit components are necessarily separate physical entities. For example, a reference to a first circuit component being “connected to” a second circuit component could encompass: (A) a scenario in which a physical structure that serves as a terminal of the first circuit component is in direct physical contact with a physical structure that serves as a terminal of the second circuit; (B) a scenario in which a physical structure that serves as a terminal of the first circuit component is in direct physical contact with an electrical conductor (e.g., a wiring line) that is itself in direct physical contact with a physical structure that serves as a terminal of the second circuit; (C) a scenario in which the same physical structure that serves as a terminal of the first circuit component also serves as a terminal of the second circuit component; etc.
When reference is made herein or in the appended claims to a first component being “connected between” second and third component, this means that two opposing terminals of the first component are connected to the second component and to the third component, respectively. In particular, when reference is made herein or in the appended claims to a transistor being “connected between” two elements, this means that a source terminal of the transistor (also referred to as a source electrode, source region, source, etc.) is connected to one of the two elements, and a drain terminal of the transistor (also referred to as a drain electrode, drain region, drain, etc.) is connected to the other one of the two elements.
When reference is made herein or in the appended claims to a number of circuit components being “connected in series between” a first element and a second element, this means that the number of circuit components are connected end-to-end in a series, in the same order that they are recited, and that the first circuit component of the series is connected to the first element and the last circuit component of the series is connected to the second element. For example, “A, B, and C are connected in series between D and E” means that D is connected to A, A is connected to B, B is connected to C, and C is connected to E, which may be graphically represented as D-{A-B-C}-E where the dashes (“-”) indicate connections and the braces (“{ }”) indicate the series.
As used herein, applying a given voltage difference “across” a circuit component (such as the memristor M) means applying voltages to two terminals of the circuit component such that the larger of the applied voltages minus the smaller of the applied voltages equals the given voltage difference. A polarity with which the voltage difference is applied across the circuit component corresponds to the direction in which current would tend to flow through the circuit component (if allowed to do so) as a result of the applied voltage difference, which will depend upon which terminal of the circuit component has the larger of the applied voltages. In particular, when the given voltage difference is applied across the circuit component by applying the larger of the applied voltages to a first terminal of the circuit component and the smaller of the applied voltages to a second terminal of the circuit component, this may be referred to as applying the given voltage difference with a first polarity; when the given voltage difference is applied across the circuit component by applying the larger of the applied voltages to the second terminal and the smaller of the applied voltages to a first terminal, this may be referred to as applying the given voltage difference with a second polarity.
References herein to a resistance RLRS that is associated with a low resistance state of the memristor M should not be interpreted to imply that each memristor M is set precisely to the value RLRS each time they are set to the low resistance state. In certain examples, the low resistance state of the memristor M might encompass a range of resistance values rather than just one specific resistance value. In particular, the resistance of the memristor M is not necessarily the exact same value each time that the memristor M is set to the low resistance state. This may occur, for example, as a result of inevitable manufacturing variances between memristors M as well as inventible variances in the voltages that are applied for setting/resetting operations. Instead, the resistance RLRS is a particular resistance value that is associated with the range of resistance values that comprises the low resistance state of the memristor M.
Similarly, references herein to a resistance RHRS that is associated with a high resistance state of the memristor M should not be interpreted to imply that each memristor M is set precisely to the value RHRS each time they are reset to the high resistance state. In certain examples, the high resistance state of the memristor M might encompass a range of resistance values rather than just one specific resistance value. In particular, the resistance of the memristor M is not necessarily the exact same value each time that the memristor M is reset to the high resistance state. Instead, the resistance RHRS is a particular resistance value that is associated with the range of resistance values that comprises the high resistance state of the memristor M.
The resistances RLRS and RHRS may be determined by a variety of methods. For example, a range of resistance values for the low resistance state and a range of resistance values for the high resistance state may be determined by repeatedly setting and resetting the memristor M and measuring its resistance each time and determining the ranges of values based on these measurements, and then the resistances RLRS and RHRS may be determined to be particular values within their corresponding ranges. For example, the resistance RLRS may be the highest value of the range of low resistance state values and the resistance RHRS may be the lowest value of the range of high resistance state values. In certain examples, the resistances RLRS and RHRS may be determined to be statistical aggregations of measured resistance values of the memristor M (or of multiple memristors M) in the set and reset states, respectively; for example, RLRS and RHRS may be determined to be the average, the median, the mode, etc., of their respectively corresponding measured resistance values. In certain examples, the memristor(s) M may be repeatedly set and reset and their resistances measured, and RLRS and RHRS may be determined to be the lowest and the highest resistance values, respectively, that are measured in this process.
In the description above, the transistors of the bit cell were described as all being NFET transistors, but it should be understood that all PFET transistors could be used instead. In such a case, the voltages that result in the transistors turning on and off would be reversed (low gate-source voltage turning the PFET on and high gate-source voltage turning the PFET off), and therefore voltages applied during the search operations may be reversed to preserve the desired search results.
Throughout this disclosure and in the appended claims, occasionally reference may be made to “a number” of items. Such references to “a number” mean any integer greater than or equal to one. When “a number” is used in this way, the word describing the item(s) may be written with the pluralized “s” for grammatical consistency, but this does not necessarily mean that multiple items are being referred to. Thus, for example, “a number of comparators” could encompass both one comparator and multiple comparators.
While the above disclosure has been shown and described with reference to the foregoing examples, it should be understood that other forms, details, and implementations may be made without departing from the spirit and scope of this disclosure.

Claims (18)

What is claimed is:
1. A content addressable memory comprising:
a bit cell comprising:
a memristor and a switching transistor that are connected in series between a first data line and a second data line; and
a match-line transistor connected between a match line and a rail,
wherein a gate of the match-line transistor is connected to a common node of the memristor and the switching transistor, and
the switching transistor is sized such that its channel resistance when on is between a resistance associated with a low-resistance state of the memristor and a resistance associated with a high-resistance state of the memristor.
2. The content addressable memory of claim 1, further comprising:
control circuitry that is to search the bit cell based on a first search criterion, search the bit cell based on a second search criterion, and search the bit cell based on a wildcard search criterion.
3. The content addressable memory of claim 1, further comprising:
control circuitry that is to:
search the bit cell based on a first search criterion by pre-charging the match line, applying a first voltage to the first data line, applying a ground voltage to the second data line, and applying a third voltage to the rail; and
search the bit cell based on a second search criterion by pre-charging the match line, applying the ground voltage to the first data line, applying the first voltage to the second data line, and applying a fourth voltage to the rail.
4. The content addressable memory of claim 3,
wherein the third voltage is set such that, when the bit cell is searched based on the first search criterion:
the match-line transistor is off when the value stored in the bit cell is a match, and
the match-line transistor is on when the value stored in the bit cell is a miss.
5. The content addressable memory of claim 4,
wherein the third voltage is set so as to be greater than a value resulting from multiplying the first voltage by
R T R H + R T
and subtracting a threshold voltage of the match line transistor, where RH is a resistance associated with a high resistance state of the memristor, and RT is a channel resistance of the switching transistor when on.
6. The content addressable memory of claim 5,
wherein the third voltage is set so as to be less than a value resulting from multiplying the first voltage by
R T R T + R L
and subtracting the threshold voltage of the match line transistor, where RL is a resistance associated with a low resistance state of the memristor.
7. The content addressable memory of claim 3,
wherein the fourth voltage is set such that, when the bit cell is searched based on the second search criterion:
the match-line transistor is off when the value stored in the bit cell is a match, and
the match-line transistor is on when the value stored in the bit cell is a miss.
8. The content addressable memory of claim 7,
wherein the fourth voltage is set so as to be greater than a value resulting from multiplying the first voltage by
R L R L + R T
and subtracting a threshold voltage of the match line transistor, where RL is a resistance associated with a low resistance state of the memristor, and RT is a channel resistance of the switching transistor when on.
9. The content addressable memory of claim 8,
wherein the fourth voltage is set so as to be less than a value resulting from multiplying the first voltage by
R H R T + R H
and subtracting the threshold voltage of the match line transistor, where RH is a resistance associated with a high resistance state of the memristor.
10. The content addressable memory of claim 3,
wherein the control circuitry is to search the bit cell based on a wildcard search criterion by applying the ground voltage to the first data line, the second data line, and the rail.
11. A content addressable memory comprising:
a plurality of bit cells arrayed in rows and columns;
a plurality of match lines, each corresponding to N bit cells that are in the same row as one another, where N is a word size of the content addressable memory;
a plurality of first data lines, each corresponding to one of the columns of bit cells;
a plurality of second data lines, each corresponding to one of the columns of bit cells;
a plurality of rails, each corresponding to one of the columns of bit cells;
wherein each of the plurality of bit cells includes:
a memristor and a switching transistor that are connected in series between the corresponding one of the plurality of first data lines and the corresponding one of the plurality of second data lines, where the switching transistor is to control whether the bit cell is selected for a writing operation and whether the bit cell is selected for a reading; and
a match-line transistor connected between the corresponding one of the plurality of match lines and the corresponding one of the plurality of rails, a gate of the match-line transistor being connected to a common node of the memristor and switching transistor.
12. The content addressable memory of claim 11,
further comprising control circuitry that is to write data to the plurality of bit cells, read data stored in the plurality of bit cells, and search for a word stored in the plurality of bit cells by applying voltages to the plurality of match lines, the plurality of first data lines, the plurality of second data lines, and the plurality of rails,
wherein the control circuitry is to search for a word stored in the plurality of bit cells on the basis of a first search criterion, on the basis of a second criterion, and on the basis of a wildcard criterion.
13. The content addressable memory of claim 12,
wherein the control circuitry is to, in searching for a word stored in the plurality of bit cells, search all of the plurality of bit cells simultaneously.
14. The content addressable memory of claim 11,
wherein, for each of the plurality of bit cells, the switching transistor is sized such that its channel resistance when on is between a resistance associated with a low-resistance state of the memristor and a resistance associated with a high-resistance state of the memristor.
15. The content addressable memory of claim 11, further comprising:
control circuitry that is to write data to the plurality of bit cells, read stored data from the plurality of bit cells, and search for a word stored in the plurality of bit cells,
wherein the control circuitry is to, in searching for a word stored in the plurality of bit cells, search a given column of bit cells out of the plurality of columns based on a first search criterion by: pre-charging each of the match lines corresponding to the given column of bit cells, applying a first voltage to the corresponding one of the first data lines, applying a ground voltage to the corresponding one of the second data lines, and applying a third voltage to the corresponding one of the rails; and
the control circuitry is to, in searching for a word stored in the plurality of bit cells, search a second given column of bit cells out of the plurality of columns based on a second search criterion by: pre-charging each of the match lines corresponding to the second given column of bit cells, applying the ground voltage to the corresponding one of the first data lines, applying the first voltage to the corresponding one of the second data lines, and applying a fourth voltage to the corresponding one of the rails.
16. An electronic device, comprising:
a content addressable memory (CAM) that comprises a plurality of bit cells that each includes a resistive divider whose output voltage controls a match-line transistor, the resistive divider comprising a memristor and a switching transistor connected in series between first and second data lines,
wherein the CAM is to search a given bit cell of the plurality of bit cells based on a first search criterion by applying a first voltage difference across the resistive divider with a first polarity,
the CAM is to search the given bit cell based on a second search criterion by applying the first voltage difference across the resistive divider with a second polarity, and
for each of the plurality of bit cells, the switching transistor is sized such that its channel resistance when on is between a resistance associated with a low resistance state of the memristor and a resistance associated with a high resistance state of the memristor.
17. The electronic device of claim 16,
wherein the CAM is to search the given bit cell based on a wildcard criterion by applying a ground voltage to both input terminals of the resistive divider.
18. The content addressable memory of claim 11, further comprising:
a plurality of word lines, each corresponding to N bit cells that are in the same row as one another, wherein each of the switching transistors has a gate electrode connected to a corresponding one of the plurality of word lines; and
control circuitry that is to select a group of N bit cells for a writing operation by applying a control signal to one of the plurality of word lines corresponding to the group of N bit cells.
US15/216,011 2016-07-21 2016-07-21 Content addressable memories Active US9721661B1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US15/216,011 US9721661B1 (en) 2016-07-21 2016-07-21 Content addressable memories
KR1020170092234A KR20180011016A (en) 2016-07-21 2017-07-20 Memristor based content addressable memories

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US15/216,011 US9721661B1 (en) 2016-07-21 2016-07-21 Content addressable memories

Publications (1)

Publication Number Publication Date
US9721661B1 true US9721661B1 (en) 2017-08-01

Family

ID=59382797

Family Applications (1)

Application Number Title Priority Date Filing Date
US15/216,011 Active US9721661B1 (en) 2016-07-21 2016-07-21 Content addressable memories

Country Status (2)

Country Link
US (1) US9721661B1 (en)
KR (1) KR20180011016A (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019108401A1 (en) 2017-11-30 2019-06-06 Micron Technology, Inc. Comparing input data to stored data
JP2020053532A (en) * 2018-09-26 2020-04-02 Tdk株式会社 Memristor circuit, memristor control system, analog product-sum computing unit and neuromorphic device
US20210021620A1 (en) * 2018-04-30 2021-01-21 Hewlett Packard Enterprise Development Lp Updating regular expression pattern set in ternary content-addressable memory
US10998047B1 (en) * 2020-01-15 2021-05-04 Hewlett Packard Enterprise Development Lp Methods and systems for an analog CAM with fuzzy search
CN113658627A (en) * 2021-07-26 2021-11-16 安徽大学 10T4R unit circuit capable of distinguishing resistance state crossing
US20220375536A1 (en) * 2021-05-20 2022-11-24 Hewlett Packard Enterprise Development Lp Analog content addressable memory for storing and searching arbitrary segments of ranges
US11561607B2 (en) 2020-10-30 2023-01-24 Hewlett Packard Enterprise Development Lp Accelerating constrained, flexible, and optimizable rule look-ups in hardware
US11615827B2 (en) 2020-10-15 2023-03-28 Hewlett Packard Enterprise Development Lp Hardware accelerator with analog-content addressable memory (a-CAM) for decision tree computation
US11622731B2 (en) * 2016-10-18 2023-04-11 University Of Southampton Method and system for detecting an event and determine information about it (like its strength) using resistive state changes of a memristor
US20230137079A1 (en) * 2021-10-29 2023-05-04 Hewlett Packard Enterprise Development Lp Iterative programming of analog content addressable memory

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6606263B1 (en) 2002-04-19 2003-08-12 Taiwan Semiconductor Manufacturing Company Non-disturbing programming scheme for magnetic RAM
US8179739B2 (en) 2007-08-10 2012-05-15 Renesas Electronics Corporation Semiconductor device and its manufacturing method
US20140355330A1 (en) 2012-01-01 2014-12-04 Tohoku University Integrated circuit
WO2015085093A1 (en) 2013-12-06 2015-06-11 Rambus Inc. 2t-1r architecture for resistive ram
US9087572B2 (en) * 2012-11-29 2015-07-21 Rambus Inc. Content addressable memory
US20160203868A1 (en) * 2015-01-14 2016-07-14 Flashsilicon Incorporation Configurable non-volatile content addressable memory
US9431106B1 (en) * 2015-06-09 2016-08-30 Freescale Semiconductor, Inc. Ternary content addressable memory (TCAM) with magnetic tunnel junction (MTJ) devices

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6606263B1 (en) 2002-04-19 2003-08-12 Taiwan Semiconductor Manufacturing Company Non-disturbing programming scheme for magnetic RAM
US8179739B2 (en) 2007-08-10 2012-05-15 Renesas Electronics Corporation Semiconductor device and its manufacturing method
US20140355330A1 (en) 2012-01-01 2014-12-04 Tohoku University Integrated circuit
US9087572B2 (en) * 2012-11-29 2015-07-21 Rambus Inc. Content addressable memory
WO2015085093A1 (en) 2013-12-06 2015-06-11 Rambus Inc. 2t-1r architecture for resistive ram
US20160203868A1 (en) * 2015-01-14 2016-07-14 Flashsilicon Incorporation Configurable non-volatile content addressable memory
US9431106B1 (en) * 2015-06-09 2016-08-30 Freescale Semiconductor, Inc. Ternary content addressable memory (TCAM) with magnetic tunnel junction (MTJ) devices

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Pilin Junsangsri and Fabrizio Lombardi, "A Memristor-Based TCAM (Ternary Content Addressable Memory) Cell: Design and Evaluation," GLSVLSI'12, May 3-4, 2012, Salt Lake City, Utah, USA, pp. 311-314, ACM.
Qing Guo et al., "AC-DIMM: Associative Computing with STT-MRAM," ISCA'13, Tel-Aviv, Israel, Jun. 2013, pp. 189-200, ACM.

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11622731B2 (en) * 2016-10-18 2023-04-11 University Of Southampton Method and system for detecting an event and determine information about it (like its strength) using resistive state changes of a memristor
EP3718108A4 (en) * 2017-11-30 2021-08-18 Micron Technology, Inc. Comparing input data to stored data
US11430511B2 (en) 2017-11-30 2022-08-30 Micron Technology, Inc. Comparing input data to stored data
WO2019108401A1 (en) 2017-11-30 2019-06-06 Micron Technology, Inc. Comparing input data to stored data
US20210021620A1 (en) * 2018-04-30 2021-01-21 Hewlett Packard Enterprise Development Lp Updating regular expression pattern set in ternary content-addressable memory
JP2020053532A (en) * 2018-09-26 2020-04-02 Tdk株式会社 Memristor circuit, memristor control system, analog product-sum computing unit and neuromorphic device
US10998047B1 (en) * 2020-01-15 2021-05-04 Hewlett Packard Enterprise Development Lp Methods and systems for an analog CAM with fuzzy search
US11694749B2 (en) 2020-01-15 2023-07-04 Hewlett Packard Enterprise Development Lp Methods and systems for an analog CAM with fuzzy search
US11615827B2 (en) 2020-10-15 2023-03-28 Hewlett Packard Enterprise Development Lp Hardware accelerator with analog-content addressable memory (a-CAM) for decision tree computation
US11561607B2 (en) 2020-10-30 2023-01-24 Hewlett Packard Enterprise Development Lp Accelerating constrained, flexible, and optimizable rule look-ups in hardware
US11551771B2 (en) * 2021-05-20 2023-01-10 Hewlett Packard Enterprise Development Lp Analog content addressable memory for storing and searching arbitrary segments of ranges
US20220375536A1 (en) * 2021-05-20 2022-11-24 Hewlett Packard Enterprise Development Lp Analog content addressable memory for storing and searching arbitrary segments of ranges
CN113658627A (en) * 2021-07-26 2021-11-16 安徽大学 10T4R unit circuit capable of distinguishing resistance state crossing
CN113658627B (en) * 2021-07-26 2024-03-29 安徽大学 10T4R unit circuit capable of distinguishing resistance state intersection
US20230137079A1 (en) * 2021-10-29 2023-05-04 Hewlett Packard Enterprise Development Lp Iterative programming of analog content addressable memory
US11783907B2 (en) * 2021-10-29 2023-10-10 Hewlett Packard Enterprise Development Lp Iterative programming of analog content addressable memory

Also Published As

Publication number Publication date
KR20180011016A (en) 2018-01-31

Similar Documents

Publication Publication Date Title
US9721661B1 (en) Content addressable memories
US9847132B1 (en) Ternary content addressable memories
US9934857B2 (en) Ternary content addressable memories having a bit cell with memristors and serially connected match-line transistors
US7855923B2 (en) Write current compensation using word line boosting circuitry
US9672912B2 (en) Semiconductor integrated circuit device with reduced power consumption
US10896731B1 (en) Content addressable memory circuits with threshold switching memristors
US9087572B2 (en) Content addressable memory
US10560085B2 (en) Apparatuses for reducing off state leakage currents
US10847224B1 (en) Low power and area ternary content addressable memory circuit
US7307861B1 (en) Content addressable memory (CAM) cell bit line architecture
JP7216436B2 (en) neural network circuit device
EP3136398B1 (en) Ternary content addressable memory (tcam) with programmable resistive elements
US10395710B1 (en) Magnetic memory emulating dynamic random access memory (DRAM)
US9312006B2 (en) Non-volatile ternary content-addressable memory with resistive memory device
US9761309B2 (en) Sensing circuit for resistive memory array
US6101140A (en) Sense amplifier driver circuit for supplying a reduced driving voltage to sense amplifier
US10783936B2 (en) Reading device and logic device
US20230013081A1 (en) Driving method of synapse circuit
US11631459B2 (en) Dual compare ternary content addressable memory
US10395719B2 (en) Memory device driving matching lines according to priority
US9842991B2 (en) Memory cell with redundant carbon nanotube
KR102505089B1 (en) NAND Type Content AddressaBLBe Memory Cell Based on Ferroelectric Element and Content AddressaBLBe Memory Including Same
US7046537B2 (en) Reduced signal swing in bit lines in a CAM
KR20230162331A (en) Content Addressable Memory Cell Based on Ferroelectric Transistor and Content Addressable Memory Including Same
US20240021243A1 (en) Sense circuit and high-speed memory structure incorporating the sense circiut

Legal Events

Date Code Title Description
AS Assignment

Owner name: HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP, TEXAS

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BUCHANAN, BRENT;ZHENG, LE;STRACHAN, JOHN PAUL;SIGNING DATES FROM 20160715 TO 20160720;REEL/FRAME:039212/0696

STCF Information on status: patent grant

Free format text: PATENTED CASE

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 4